Long-term survival after childhood acute lymphoblastic leukaemia: population-based trends in cure and relapse by clinical characteristics.

'Cure models' offer additional information to traditional epidemiological approaches to assess survival for cancer patients by simultaneously estimating the proportion cured and the survival of those 'uncured'. The proportion cured is a summary of long-term survival while the median survival time of the uncured provides important information on those who are not long-term survivors. Population-based trends in the cure proportion and survival of the uncured for childhood acute lymphoblastic leukaemia (ALL) by clinical prognostic risk factors were estimated using flexible parametric cure models, based on overall survival and event-free survival. Children aged 1-17 years diagnosed between 1990 and 2011 in Yorkshire, UK, were included (n = 492). The percentage cured increased from 77% (95% confidence interval 70-84%) in 1990-1997 to 89% (84-93%) in 2003-2011, while the median survival time of the uncured decreased from 3·2 years (2·2-4·1 years) to 0·7 years (0-1·5 years). Models based on event-free survival showed a similar trend. The 5-year cumulative incidence of relapse substantially decreased from 35% in 1990-97 to 9% in 2003-2011. These results show selective improvement in survival between 1990 and 2011 with a significant reduction in the risk of relapse alongside a reduced absolute duration of survival for those destined to be uncured.